kirlik@chmsr.gatech.edu (10/01/89)
Barry Kort writes: >>In article <1882@csm9a.UUCP> fhadsell@csm9a.UUCP ( GP) writes: > I have received a flyer concerning a course by Llinas & Hall on Data > Fusion. As I read it the flyer doesn't say what Data Fusion is. Would > you please enlighten me? >> You are sitting at a large table, assembling a giant jigsaw puzzle. (Some stuff deleted) >> You will have to >> to use deductive and inductive reasoning, inferential reasoning, >> combinatorial logic, analogical and model-based reasoning, and >> diagnostic reasoning to accomplish your task. >> That is Data Fusion. I'm not exactly sure why, but I find this (type of) problem definition somewhat troublesome. Perhaps it is because it seems to smuggle the solution methodology into the statement of the problem. Deductive and inductive reasoning, inferential reasoning, analogical, model-based, and diagnostic reasoning are theory laden terms, reflective of an (in my opinion premature and naive) way of carving up the AI problem (and definitely premature if it reflects a way of carving up the psychological problem). These terms are not part of the data of experience, nor have the theories giving rise to them reached a level of pragmatic utility and uncritical acceptance that we should be using them in our problem Definitions. If this way of carving up the world is wrong, all sorts of false problems can be created. In short, I'd make the guess that solutions to the data fusion problem are possible without using all these "reasonings", and that I am definitely not using them when puting together a jigsaw puzzle, although I may be described as if I was using them. I only got up the gumption to make this posting because I see a lot of "getting the cart before the horse" in AI, and too many people armed with solutions looking for problems. I guess this will endear me to the community. Alex Kirlik UUCP: kirlik@chmsr.UUCP {backbones}!gatech!chmsr!kirlik INTERNET: kirlik@chmsr.gatech.edu
fhadsell@csm9a.UUCP ( GP) (10/01/89)
In article <2185@hydra.gatech.EDU>, kirlik@chmsr.gatech.edu writes: > > Barry Kort writes: > > (Some stuff deleted) > > >> You will have to > >> to use deductive and inductive reasoning, inferential reasoning, > >> combinatorial logic, analogical and model-based reasoning, and > >> diagnostic reasoning to accomplish your task. > > >> That is Data Fusion. > > I'm not exactly sure why, but I find this (type of) problem definition > somewhat troublesome. Perhaps it is because it seems to smuggle the > solution methodology into the statement of the problem. ......... > .... These terms are not part of the data of > experience, nor have the theories giving rise to them reached a > level of pragmatic utility and uncritical acceptance that we should > be using them in our problem Definitions. ........ > . . . . . . . Alex Kirlik substantiates my worst fears. We AI'ers have been guilty of the careless definition of terms for a long long time, and these sins are beginning to catch up with us. Our customers; i.e., those few that remain, are much better informed than they were. They speak knowledgeably of the various parts of the ill-defined body which is AI. In mineral exploration we speak of the integrated approach when we employ more than one discipline, such as geology, geophysics, and/or geochemistry. Each of these differ in their use of the different flavors of logic and reasoning. Integrated exploration is rapidly becoming simply exploration. In other words the real world demands the interdisciplinary approach now. All of the three geos use all of the commonly accepted methods of AI; e.g., expert systems, artificial neuralnets, hypermedia, natural language, etc., but I fear that if I were to speak of datafusion in the exploration business I would now be ridiculed. We have enough terms in AI. Let's now demonstrate, on a grander scale, the utility of our science. -- fhadsell@csm9a.colorado.edu Frank Hadsell Professor of Geophysics, Colorado School of Mines, Golden, Colorado 80401 (303) 273-3456